Crafting a Deliverable - DATA 171 Lesson 10

Comprehensive vocabulary flashcards covering the components, planning stages, evaluation criteria (CRAAP test), reference tools, statistical methods, and structural requirements for data science deliverables in DATA 171 Lesson 10.

Final Deliverable

The formal document or presentation whose ultimate purpose is to communicate analytical findings and insights to stakeholders who formulate policy or strategy, typically spanning $1000$ to $7000$ words ($2$ to $14$ pages).

Research Question

The foundational inquiry that defines the objective of an analytical report and must be refined to control for confounding factors (such as education, experience, and productivity when evaluating wage disparities).

Literature Search

The systematic academic process of examining prior published studies to determine how other researchers have investigated the same or related research questions.

Background Research

The industry counterpart to an academic literature search, focusing on existing industry information, prior inquiries, and related solutions.

CRAAP Test

A source evaluation framework developed by California State University, Chico based on five criteria: Currency, Relevance, Authority, Accuracy, and Purpose.

Currency (CRAAP Test)

The criterion assessing the timeliness of information, including publication date, updates or revisions, topic requirements, and functional hyperlinks.

Relevance (CRAAP Test)

The criterion assessing how well information answers the research question, fits the target audience, matches the appropriate academic depth, and compares against alternative sources.

Authority (CRAAP Test)

The criterion evaluating the author, publisher, sponsor, or organization, their credentials, contact information, and top-level domain credibility (e.g., .org, .edu, .gov).

Accuracy (CRAAP Test)

The criterion assessing the truthfulness and reliability of content, verifying evidence, referee/peer-review status, unbiased tone, and absence of errors.

Purpose (CRAAP Test)

The criterion determining why information exists (to inform, teach, sell, entertain, or persuade) and identifying potential political, ideological, cultural, or personal biases.

Zotero

A reference management software recommended by Murtaza Haider for collecting and managing citations and bibliographic references.

Evernote

A note-taking application recommended by Murtaza Haider for recording and organizing detailed notes on collected reference materials.

BibTeX

A reference management format and tool used to store, manage, and format citations when composing technical documents in LaTeX.

Data Acquisition Techniques

Methods used to gather missing data when existing secondary sources are insufficient, specifically web scraping, surveys, and focus groups.

Analysis of Variance (ANOVA)

A statistical method used when a researcher wants to evaluate and analyze the effects of categorical factors across groups.

Regression Analysis

A statistical technique used to model the relationship between dependent and independent variables with the aim of making predictions.

Standard Deliverable Structure

The core organization of an analytics report: Cover Page, Table of Contents, Introduction, Literature Review / Related Work, Methodology, Results, Discussion, Conclusion, and References.

Haider's 7-Item Checklist

A review framework verifying that a report: defines reader benefits, clarifies aims, explains significance, cites literature, addresses practical usefulness, identifies future developments, and maintains logical structure.

Final Deliverable

The formal document or presentation created to communicate analytical insights, methodologies, and actionable recommendations to stakeholders and decision-makers who formulate policy or strategy, typically spanning $1000$ to $7000$ words ($2$ to $14$ pages).

Research Question Refinement

The analytical process of narrowing and specifying an initial inquiry to account and control for confounding variables (such as education, experience, and productivity when evaluating wage disparities) to ensure the findings answer the precise policy or business issue.

Literature Search

The systematic academic examination of published scholarly articles, peer-reviewed journals, and institutional research to investigate how prior investigators approached identical or closely related research inquiries.

Background Research

The industry counterpart to an academic literature search, focused on investigating prior corporate inquiries, existing commercial data, proprietary internal records, and related business case studies.

CRAAP Test

A critical source evaluation framework developed by California State University, Chico to assess information reliability using five core criteria: Currency, Relevance, Authority, Accuracy, and Purpose.

Currency (CRAAP Test)

The evaluation dimension assessing the timeliness of information, including initial publication dates, subsequent updates or revisions, topic-specific recency requirements, and functional hyperlinks.

Relevance (CRAAP Test)

The evaluation dimension assessing how appropriately source content addresses the specific research question, matches the intended audience, provides appropriate depth, and compares in quality to alternative sources.

Authority (CRAAP Test)

The evaluation dimension examining the author, publisher, sponsor, or organization, verifying their credentials, organizational affiliations, contact details, and top-level domain credibility (e.g., .edu, .gov, or .org).

Accuracy (CRAAP Test)

The evaluation dimension verifying the truthfulness and reliability of source content through empirical evidence, peer-review or referee status, neutral language, and the complete absence of factual or typographical errors.

Purpose (CRAAP Test)

The evaluation dimension identifying why the information exists (to inform, teach, sell, entertain, or persuade) and detecting any underlying political, commercial, ideological, cultural, or personal biases.

Zotero and Evernote Workflow

A reference management pairing recommended by Murtaza Haider where one tool collects, stores, and organizes bibliographic citations, while the companion application manages detailed qualitative notes on those reference materials.

BibTeX

A citation management tool and plain-text file format utilized within LaTeX technical typesetting environments to manage, structure, and render bibliographies and academic references.

Primary Data Acquisition Techniques

Empirical data gathering methods utilized when existing secondary datasets are insufficient or unavailable to answer the research inquiry, specifically web scraping, structured surveys, and qualitative focus groups.

Standard Analytics Deliverable Structure

The formal organizational hierarchy of a data science report: Cover Page, Table of Contents, Introduction, Literature Review / Related Work, Methodology, Results, Discussion, Conclusion, and References.

Haider's Seven-Item Checklist

A quality-assurance framework confirming that a report clarifies reader benefits upfront, articulates clear aims, highlights contribution significance, contextualizes findings in prior literature, demonstrates practical usefulness, outlines future developments, and maintains logical structural flow.

Grammar of Graphics

A principled visualization framework implemented in ggplot2 that abstracts any graphical plot into layered components: Data, Aesthetic mappings, Geometries, Statistics, Scales, Coordinate Systems, Facets, and Themes.

Aesthetic Mappings (aes)

The visual channel bindings in ggplot2 that map variables from a dataset to graphical properties of a plot, including the $x$-axis, $y$-axis, color, fill, shape, line type, and size.

stat = "identity" in geom_bar

The parameter instruction in ggplot2 that tells a bar chart to plot explicit raw values provided directly by the $y$-aesthetic, overriding the default behavior which counts frequencies.

Jitter Plot (geom_jitter)

A modified scatter plot geometry that introduces a small amount of random positional noise along the horizontal or vertical axes to prevent overplotting and reveal dense clusters of overlapping data points.

Density Plot (geom_density)

A smoothed continuous representation of the empirical probability distribution of a continuous variable, frequently paired with alpha transparency to compare overlapping distributions across categorical groups.

Position Dodging (position = "dodge")

A positioning parameter in ggplot2 that places graphical elements, such as grouped histogram bins or bar charts, side-by-side rather than stacking or overlapping them on top of one another.

Anscombe's Quartet

A set of four synthetic datasets that exhibit identical numerical summary statistics (such as means, sample variances, and correlation coefficients) but show completely distinct patterns when graphed, demonstrating that summary statistics alone can fail to reveal true underlying data structures.

Information Visualization vs. Scientific Visualization

The distinction between visually exploring interactive representations of abstract, categorical variables to discover patterns, trends, clusters, and outliers versus rendering continuous physical variables, volumes, and geometric surfaces (such as airflow or medical CT scans).

Parallel Coordinate Plot

A multi-dimensional data visualization where each vertical, parallel axis represents a distinct quantitative or categorical attribute, and each individual observation is rendered as a polyline connecting its values across each dimension.

Tree Data vs. Network Data Visualization

The architectural distinction between visualizing hierarchical collections where each child item (except the root) links to a single parent (e.g., treemaps or organizational charts) versus visualizing relational graphs where nodes have arbitrary, unconstrained, and non-hierarchical links.

Faceting (facet_wrap vs. facet_grid)

The sub-plotting mechanism in ggplot2 where one function lays out a one-dimensional ribbon of panels wrapped across two dimensions based on a single variable, while the companion function creates a two-dimensional grid of panels defined by explicit row and column formulas.

Coordinate Zooming vs. Limit Deletion

The functional difference between using coord_cartesian() to visually magnify a plot area without removing data points, versus applying xlim() or ylim(), which deletes out-of-bounds data and alters statistical summary layers such as regression best-fit lines.

Theme Customization in ggplot2

The non-data visual formatting layer in ggplot2 controlled through preset functions (such as theme_classic(), theme_bw(), or external packages like ggthemes) or customized via theme() using element_text(), element_line(), element_rect(), and element_blank().

Unit Normal Distribution

A standard continuous probability distribution with a mean of $\mu = 0$ and a standard deviation and variance of $\sigma = 1$, possessing a total area under the probability density curve equal to unity ($1$).

Standard Score ($Z$-Score)

A standardized statistical measure representing the signed number of standard deviations an individual raw score deviates from the distribution mean, calculated as $Z = \frac{X_i - \bar{X}}{s}$.

Empirical Rule ($68-95-99.7$ Rule)

The normal distribution property stating that approximately $68.26\%$ of values fall within $\pm 1\sigma$ of the mean, approximately $95.44\%$ fall within $\pm 2\sigma$, and approximately $99.72\%$ fall within $\pm 3\sigma$.

Skewness

A measure of the directional asymmetry of a distribution, classified as positive skew when the tail tapers toward higher values (mean greater than median) and negative skew when the tail tapers toward lower values (mean less than median).

Kurtosis

A statistical descriptor characterizing the sharpness or peakedness of a frequency distribution relative to a standard normal curve, divided into leptokurtic (sharp peak, heavy tails), mesokurtic (normal peak), and platykurtic (flat peak, light tails).

Null Hypothesis ($H_0$) vs. Alternative Hypothesis ($H_1$)

The formal dichotomy in statistical testing where the baseline model posits no effect, zero difference, or no association, whereas the investigator's competing hypothesis posits a statistically significant difference or directional effect.

Level of Significance ($\alpha$-Level)

The predetermined probability cutoff (commonly set at $\alpha = 0.05$ or $\alpha = 0.01$) against which a test's $p$-value is evaluated to determine whether the null hypothesis should be rejected.

Type I Error vs. Type II Error

The two primary decision failures in hypothesis testing: the former occurs when a researcher rejects a true null hypothesis (false positive), while the latter occurs when a researcher fails to reject a false null hypothesis (false negative).

Sampling Distribution

The theoretical probability distribution of a given sample statistic (such as a sample mean) calculated from all conceivable random samples of a designated size $N$ drawn from a broader population.

Standard Error of the Mean ($SE$)

The standard deviation of the sampling distribution of the sample mean, calculated as $SE = \frac{s}{\sqrt{N - 1}}$ when estimating from a sample, or $SE = \frac{\sigma}{\sqrt{N}}$ when the true population standard deviation is known.

Confidence Interval

A computed range centered around a sample statistic with an associated probability (such as $95\%$) that specifies how frequently the true, unobserved population parameter is bounded by the calculated interval across repeated sampling.

Single-Sample $Z$-Test

A parametric test of significance evaluating whether an observed sample mean differs from a known population mean, requiring an assumed normal population distribution, a known population standard deviation $\sigma$, and typically a sample size $N > 30$.

Student's $t$-Test

A parametric significance test introduced by William Sealy Gosset to evaluate sample means when the true population standard deviation is unknown and estimated using the sample standard deviation $s$, especially appropriate for small sample sizes ($N < 30$).

Two-Sample Independent $t$-Test

A parametric test used to compare the difference between the means of two independent, unrelated groups with unequal variances, computed via $t = \frac{\bar{X}_1 - \bar{X}_2}{\sqrt{\frac{s_1^2}{N_1} + \frac{s_2^2}{N_2}}}$.

Mann-Whitney $U$ Test

A non-parametric statistical procedure that evaluates whether two independent groups differ significantly across ranked ordinal, interval, or ratio data without requiring an assumption of normal population distributions.

Chi-Square Test

A non-parametric test of significance utilized for nominal and categorical level data to analyze whether observed frequency counts across discrete categories differ significantly from expected theoretical frequencies.